Anthropic keys (sk-ant-) in client bundles — detection and rotation
sk-ant- strings in served JS are billable Claude API access. Find them in seconds — free scan, no signup.
Long prefixes make Anthropic keys distinctive — and their exposure identical in consequence to any other live-model credential.
What an Anthropic API key is
sk-ant-api…/sk-ant-admin… authenticate Claude API calls with version segments embedded.
sk-ant-api03-FAKE0000000000000000000000000000000000000000000000The prefix is the claim: scanners and attackers alike identify the format before they know anything else about it, which is why recognition starts at the first characters rather than the last.
What it grants
Within Anthropic’s platform:
- Messages/inference billed to your console org.
- Admin variants add workspace management reach.
How it ends up in a bundle
Three arrival routes cover nearly every case we see:
- Agent-written chat widgets calling directly.
- Copied quickstart code retaining sample structure but real keys.
Does it belong in a browser?
Server-only; proxy pattern identical to OpenAI’s.
Rotate it
console.anthropic.com/settings/keys → revoke/create/update-env.
Find it in seconds
Open DevTools on the deployed site and search the built assets for sk-ant-. If the search hits, the credential shipped; if it does not, check the chunks loaded on authenticated or interactive views, not just the landing page — the calling code often sits behind a route.
The faster path is to let a machine do the fetching. KeyDrift downloads the same JavaScript a visitor gets — HTML, every referenced chunk including ones named only in the route manifest, and the server-streamed data frameworks inline into the document — and reports credentials with a masked prefix, a fingerprint, and the exact file they live in. Paste your deployed URL into the scanner; no account needed.
Make sure it stays gone
It bears saying because it happens constantly: the fix holds until the next prompt that needs the query to return rows. Drift monitoring exists for precisely this — it diffs consecutive scans and pages you when a previously resolved finding reappears, naming the regression as a regression rather than repeating the first alert.
Why this failure class persists
It helps to name the economics honestly. Fixing this class of leak costs minutes when caught at deploy time and days when caught at invoice time, because by then the credential has been harvested, validated, resold or drained — often all four. Detection latency is the entire game, which is why the monitoring half of KeyDrift exists alongside the scanning half.
How KeyDrift reports this exact finding
When KeyDrift finds this on your deployment, the report shows a masked value (first 8 and last 4 characters only), a salted fingerprint for tracking, the exact chunk filename carrying it, and a severity with written rationale. Public-by-design neighbours — anon keys, publishable keys, Firebase web constants — appear as informational context rather than noise, because knowing what should be there is what makes the real findings credible.
Manual check, step by step
A five-minute version you can run anywhere: view-source on the landing page, copy every src= script URL, fetch each and search the results for sk-ant-. It misses manifest-only chunks and streamed payloads — which is precisely the gap between "I checked" and "it is clean" — but it catches the loud majority and builds the pattern-recognition that makes scanner output legible.
Close the loop with monitoring
Monitoring closes the loop that one-time verification leaves open. A scheduled scan refetches everything, diffs against history, and fires only on transitions: created, regressed, resolved. Regression alerts matter most here — they fire when a previously fixed finding returns, which in agent-era codebases is less a possibility than a schedule.
Common questions
Streaming via proxy hard?
Same passthrough approach as other providers.
Workspaces limit blast radius?
They scope access; publication still demands immediate rotation.
Run a free scan at keydrift.dev/scan — paste a URL or the bundle source itself, no account. Findings arrive masked, with the exact chunk they live in.